Search AI Products and News

AI startup Luma recently announced on X that it has open-sourced a new image model pretraining technique called Inductive Moment Matching (IMM). This breakthrough technology has garnered significant attention for its efficiency and stability, considered a major advancement in the generative AI field. According to X user linqi_zhou, IMM is a novel generative paradigm capable of stable training from scratch with a single model and single objective, surpassing existing methods in both sampling efficiency and sample quality.

Against the backdrop of rapid advancements in artificial intelligence, research and applications of AI large models are booming. Recent developments from various companies and institutions in this field have garnered significant attention. Firstly, Nvidia's price-to-earnings ratio before the launch of ChatGPT attracted considerable attention on Wall Street, with experts suggesting this might foreshadow a repeat of Apple's success story in 2008. Meanwhile, Yu Feng, CIO of Guotai Junan Securities, stated that the rise of AI large models will propel the securities industry into an era of "intelligent cognition".

In recent years, the programming capabilities of Artificial Intelligence (AI) have been continuously evolving, but they have not yet reached perfection. Recently, Max Woolf, a senior data scientist at BuzzFeed, discovered through experiments that if continuous prompts for 'writing better code' are provided to large language models (LLMs), AI can indeed generate higher quality code. This finding has garnered widespread attention, with renowned AI scientists emphasizing the importance of iteration and prompt design. In Woolf's experiment,

Yunzhisheng Intelligent Technology Co., Ltd. has recently launched an automatic live streaming system called 'Blue Algae AI Broadcasting', aimed at addressing multiple pain points in the live streaming industry through artificial intelligence technology. This system is based on Yunzhisheng's Shanhai Large Model and advanced human-like AI voice technology, enabling 24-hour continuous broadcasting, significantly reducing live streaming costs while enhancing interactivity and viewer experience.

The Kuaishou AI team has launched a membership service called 'KeLing AI', which provides a large model for video generation, supporting users to create videos up to 1080p resolution and 2 minutes in length, with customizable aspect ratios. This service offers two subscription options: continuous monthly and quarterly, with first purchase prices of 19 yuan and 99 yuan, which are approximately 30% and 50% of the original price respectively. The subsequent renewal prices are 58 yuan and 152 yuan, equating to 12% and 23% off the original prices, and users can cancel at any time. On July 6, the 'KeLing AI' web version officially launched, providing limited-time free services and integrating text generation functionality.

On July 11th, Kuaikan Comics announced that they are exploring fine-tuning of open-source large models to train a vertical large model for the two-dimensional (2D) domain. Kuaikan Comics utilizes large language models (LLM) and retrieval-augmented generation (RAG) technology to improve the conversion rate of work searches and inject new vitality into the search process. They have adopted a "fine-tuned large model + RAG enhancement" strategy, building internal PGC (Professional Generated Content)

The team from Peking University and the Hong Kong University of Science and Technology made a big splash with a new training method that has achieved GPT-4 level performance with an 8B-sized medical expert model. This is no small feat, and they have also introduced a new concept, "stability gap," to explain certain phenomena observed during the continuous pre-training of large language models.Image Source Note: The image is generated by AI, and the image is provided by Midjourney, an image autho

OpenAI has released the video generation model Sora, which excels at realistically simulating the physical world. Google has launched the language model Gemini 1.5 Pro, which performs close to 1.0 Ultra and can extend text length up to 10 million. The ChatBot Arena platform features a model called Mistral-Next, which has shown preliminary tests indicating it is at least a reliable model. Sora utilizes a combination of visual transformers and diffusion models, while Gemini 1.5 may leverage optimized TPU.